In accordance with an advance in digital image processing technology, an image forming apparatus such as a copier and a printer using an electrophotographic process or an ink-jet system has come to be capable of reproducing a color image at high quality and a full-color digital copier and multi-functional peripheral are commercialized.
Document images to be copied with these image forming apparatuses include documents of text, halftone photographs, line drawings, photographs (continuous gray scale area such as a photograph), or documents of those mixed and to obtain a good reproduced image, it is necessary to perform image processing suitable for the type of component of each documents.
Segmentation processing is known that performs processing of segmenting into a text area, a halftone area, and others (a photograph area and an area such as page background not segmentable into the text area or the halftone area) using information of a mask centered around a target pixel (Japanese Laid-Open Patent Publication No. 11-69150), as a method of discriminating the type of image component of the documents.
In Japanese Laid-Open Patent Publication No. 11-69150, in the first image scanning, scanning is made at high speed, a document image is read at coarse resolution in the sub-scanning direction, and a macro segmentation section performs segmentation according to image structure as viewed in perspective and outputs a segmentation class signal. Then, the document image is read at low speed and a micro segmentation section performs area segmentation, paying attention to a difference in microscopic features of the image and outputs an image area signal. With the image area signal being selected based on the segmentation class signal output from the macro segmentation section, namely, by performing suitable image area segmentation according to results of the segmentation using the difference in macro structure features, accuracy may be improved of segmentation among text, gray scale edges, and smooth gray scale parts.
In Japanese Laid-Open Patent Publication No. 7-123253, in the case of, for example, image enlarging/reducing, etc., an image area segmentation standard for segmenting the type of the image area for each pixel is adjusted according to the reading speed, with respect to the image read at different image reading speed (resolution).
Since a change in the reading speed in the sub-scanning direction causes a change in sampling interval in the sub-scanning direction, frequency component of the image and direction of an edge in a slant direction apparently change. Accordingly, for example, with the sub-scanning speed at zoom level of 100% taken as a reference speed, when a speed ratio e of the reading speed in the sub-scanning direction to the reference speed is within the range of 1/2≦e<1 (zoom level=100 to 200%), threshold is adjusted in such direction that enables the edge component in the slant direction to be easily extracted and when the above reading speed is increased to the zoom level of less than 100% (reducing), the threshold is made equal to that of the reference speed, with respect to a halftone area.
In the case of extracting the edge, when the speed ratio e is within the range of 1≦e≦2 (zoom level=50 to 100%), adjustment is made to such direction that enables an edge image to be more easily discriminated as the reading speed is decreased (so that the threshold will be lowered) and when the speed ratio is in the range less than 2/3 exceeding the zoom level of 150%, adjustment is made conversely so that the edge will be less easy to extract (Japanese Laid-Open Patent Publication Nos. 11-69150 and 7-123253).
While Japanese Laid-Open Patent Publication No. 11-69150 describes that in performing the segmentation, reading is made at coarse resolution of 200 dpi in the sub-scanning direction as opposed to the resolution of 600 dpi in the main scanning direction, increasing (e.g., three times) the reading speed (scanning speed) in the sub-scanning direction will cause image data to be thinned out and information on the image data to be lost, resulting in a decrease in the segmentation accuracy.
For example, in a pattern of lines parallel to the main scanning direction as shown in FIG. 13, when the resolution is same for both of the main scanning direction and the sub-scanning direction, results of reading may be obtained that correspond to the line pattern as shown in (A) but when the resolution is lower in the sub-scanning direction, the results of reading are obtained that appear to indicate presence of a gray scale area or light halftone instead of the line pattern as shown in (B).
Therefore, according to the method described in Japanese Laid-Open Patent Publication No. 11-69150, the accuracy is lowered of the segmentation of the image data in the sub-scanning direction and it is possible that inappropriate processing is applied.
According to the method described in Japanese Laid-Open Patent Publication No. 7-123253, it appears that erroneous discrimination may be reduced by changing the threshold at the time of performing area determination according to the reading speed (resolution) in the sub-scanning direction but it is possible that the erroneous discrimination may not be reduced depending on the method of setting the threshold.